After Spirit Aviation Holdings, Inc. and affiliates filed chapter 11 in the Southern District of New York (Case No. 25-11897 (SHL)), the wind-down estate auctioned a package labeled Deidentified Data. On 14 August 2026 Google LLC was named successful bidder at $10,000,000, with Mercor.io Corporation as $7,500,000 alternate. Some coverage treated that result as Google buying every Spirit record, including passenger bookings and frequent-flyer files, for AI training. The docket instead shows a proposed section 363 bill of sale that still needed a sale order, third-party de-identification, and later consumer-system exclusions. For AI builders, this is a documented market for failed-company operating corpora, not a rumor about scraping an airline.
What we checked
- Reconstruct the primary agreement/evidence chain and separate data licensing facts from press interpretation. confirmed. The chain is Case No. 25-11897 (SHL) (Bankr. S.D.N.Y.): Bidding Procedures Order ECF 1213, a 14 August 2026 auction, ECF 1463 with a $10,000,000 Google bill of sale, proposed order ECF 1469, Google’s 9 September response, and the 5 October ombudsman supplement. It is a section 363 sale of Deidentified Data, not a license, and it does not match coverage that Google bought all passenger files. [1][2][6][7][4]
- Google has already closed a purchase of Spirit Airlines data and begun using it to train AI models. partly confirmed. Google won the auction and said the dataset “can be helpful in improving our products and AI models,” but closing is conditioned on a bankruptcy-court sale order and a de-identification certificate. The 5 October 2026 ombudsman supplement and 8 October 2026 congressional letter still describe a proposed sale, and no entered sale order was among the documents opened. [1][4][5][7][8]
- Press accounts that Google is buying all of Spirit’s data, including passenger profiles and frequent-flyer files, match the contract. refuted. The Assets Schedule marks Customer Profiles (97,500,000), Free Spirit members (50,200,000), savers club, surveys, call recordings, and website analytics “Not Included.” Google later told the court it had also excluded PNR/transaction, refund, inflight-purchase, and Wi-Fi sales systems entirely. CNN’s claim that the purchase included frequent-flyer information is inconsistent with that schedule. [1][6][4]
What we found
- The deal is a section 363 bill of sale, not a license. ECF 1463 attaches an unsigned Bill of Sale between Google LLC and Spirit Airlines, LLC for $10,000,000, transferring listed Assets free and clear, as-is, with no contracts assumed. Delivery cannot precede court approval and a de-identification certification under a CCPA-style standard, with health-like data held to 45 C.F.R. § 164.514. A proposed sale order (ECF 1469, 17 August 2026) remained undated and unentered on the copy opened. [1][2]
- The original schedule split consumer identity systems from enterprise operations. Not Included: Customer Profiles (97,500,000 passengers), Free Spirit members (50,200,000), savers club, social posts, surveys, call recordings, and website analytics. Included: about 100 million emails, 80,000 email accounts, 500 million Teams items, OneDrive and SharePoint corpora, employee records from August 1986 (175,658), payroll (3,426,618), tax forms (148,018), time cards (1,092,000), crew pairings, competitor-pricing feeds (7,250,630,887 Infare observations), and about 30 million lines of code in 516 repositories. [1][3]
- Headlines that Google was buying “all” Spirit data, including frequent-flyer files, overstated the contract. Google’s statement, reported by CNN, Bloomberg Law, and Axios, was that it “acquired part of an enterprise dataset” useful for “products and AI models,” would “not receive any personal information,” and would take data only after third-party scrubbing. Axios published the second sentence in full. The schedule already excluded loyalty and passenger-profile databases. [4][3][5][1]
- Consumer scope narrowed after the auction; employee risk was left open. Google’s 9 September 2026 filing said the sale “is not, and never has been, a sale of Consumer Data,” and that PNR/transaction, refund, inflight-purchase, timecard, and Wi-Fi sales datasets would not be provided even in de-identified form. Ombudsman Lucy L. Thomson’s 5 October 2026 supplement recommended consumer-side approval after those exclusions, hiring of Tonic.ai, and a public no-reidentification commitment, and she said she did not investigate employee or third-party risk. [6][7]
What it means
Treat distressed 363 sales as a real channel for operational training data—mail, chat, yield systems, tickets, and source control—priced here between Mercor’s $7.5 million backup and micro1’s $12.5 million notice, with Google at $10 million. Underwrite the schedule, not the headline. Check the Personal Data definition (here, information associable with a consumer, with Deidentified Data carved out), who selects and pays the de-identification agent, whether referential integrity is preserved by design, whether the buyer may sublicense, and whether representations survive closing (here they do not, except fraud). Consumer-ombudsman clearance is not employee-confidentiality clearance. Until a signed sale order and agent certification exist, do not treat the corpus as Google’s.
Still open
- Whether Judge Sean Lane entered a sale order after 8 October 2026, and on what revised asset list.
- Whether Team Member systems, including timecards, stayed excluded after Google’s September filing.
- What Tonic.ai will certify, and whether residual re-identification of a 4,600-person crew base is measured.
- Whether Google will train general models, aviation products, or both, and whether it will onward-transfer.